Write tests for preadv2()

[metan-ucw]

We do have tests for preadv() already so the missing coverage is:

• using -1 for offset causes preadv2() to use current file offset
• RWF_NOWAIT causes the call to return EAGAIN in a case that we would have to wait for the data
• RWF_HIPRI causes kernel to do high priority read, currently only usable with O_DIRECT

[amannch]

Started working on that

using -1 for offset causes preadv2() to use current file offset

This seems to be implemented already, I'll focus on the other two points

Edit:

RWF_NOWAIT causes the call to return EAGAIN in a case that we would have to wait for the data

There's already a patch in the mailing list for that

RWF_HIPRI causes kernel to do high priority read, currently only usable with O_DIRECT

Missing the hardware to test block device polling, someone else has to work on that

[metan-ucw]

Just pushed the basic RFW_NOWAIT test. Quoting Amir on the LTP ML "There are several other ways to get EAGAIN, like DIO read/write when page cache is populated." so there is possibly a few more ways how to increase the coverage.

[coolgw]

Just pushed the basic RFW_NOWAIT test. Quoting Amir on the LTP ML "There are several other ways to get EAGAIN, like DIO read/write when page cache is populated." so there is possibly a few more ways how to increase the coverage.

Simple do DIO read can not trigger EAGAIN after my sanity check, so i think I/O should busy also.